<![CDATA[A multidimensional approach to mapping poverty in Indonesia necessitates the use of the Human Development Index (HDI), Poverty Line, and Expenditure per Capita. This study aims to evaluate the performance of two clustering algorithms with distinct paradigms the centroid-based Fuzzy C-Means (FCM) and the density-based OPTICS in profiling poverty across 501 regencies and cities. Experimental results indicate that OPTICS achieved high internal validity, with a Silhouette Score of 0.7301 and a Davies-Bouldin Index (DBI) of 0.329. Significantly, OPTICS identified 94% of the data as noise. This finding reveals a fundamental characteristic: the distribution of socio-economic data in Indonesia is highly heterogeneous and sparse, lacking inherently dense cluster structures. Conversely, FCM, employing a soft clustering approach, successfully accommodates the ambiguity of data boundaries and provides comprehensive segmentation across all regions. Despite yielding lower validity metrics (Silhouette Score 0.3894), FCM was selected as the final model because it satisfies the practical requirements of the application, which demands complete coverage mapping. This study concludes that a soft clustering approach is more applicable than density-based clustering for analyzing highly heterogeneous data such as that found in Indonesia]]>
Copyrights © 2026
